Intelligent ditch dam adaptive adjustment method based on Internet of Things

Abstract

The present invention discloses an intelligent adaptive adjustment method for ditches, dams and gates based on the Internet of Things, which relates to the field of drainage control technology. Focusing on the fault prevention and disposal of dam and gate actuators in farmland irrigation, it provides an overall method of multi-source data fusion, predictive diagnosis and automated fault tolerance; through edge computing and wavelet noise reduction technology, high-quality sensor data is acquired in real time, and potential anomalies are marked in time; then, the actuator status is analyzed in combination with historical fault records and health models, and maintenance suggestions are issued in advance; if a serious fault occurs, the local controller automatically enables secondary adjustment and emergency linkage to temporarily compensate for downstream water supply or flood diversion; finally, the central dispatching center uses global collaborative optimization and self-learning mechanisms to iteratively update the dispatching strategy to enhance system reliability and resource utilization efficiency; thereby realizing closed-loop management of dam and gate actuators and ensuring stable operation of large-scale farmland irrigation under load and outdoor environment.

Description

Technical Field

[0001] The present invention relates to the technical field of drainage control, and in particular to an intelligent ditch dam adaptive adjustment method based on the Internet of Things. Background Art

[0002] In large-scale agricultural irrigation and water conservancy operations, canal systems often require numerous gate and dam actuators (such as electric valves and gate hoists) to precisely control water flow. Because these devices operate outdoors for extended periods, their operation is subject to significant loads and wear on mechanical components and sensor systems, often impacted by sedimentation, weed growth, metal corrosion, and frequent opening and closing operations.

[0003] Especially during peak irrigation season, gates must be opened and closed repeatedly to meet downstream water demand. If the equipment becomes stuck or the opening angle deviates, reliable irrigation of downstream farmland becomes difficult to guarantee. Furthermore, gate failure can lead to flooding, seriously impacting agricultural production safety and the stability of canal infrastructure. With the increasing adoption of IoT and automation technologies in modern irrigation, achieving continuous status monitoring and agile fault handling of gate and dam actuators to ensure accurate and safe water supply to large-scale farmland is becoming an increasingly pressing engineering challenge.

[0004] In this usage scenario, one of the core technical problems is the lack of an effective prediction and intelligent handling mechanism for actuator failures. Although traditional systems can sense gate anomalies through regular inspections or simple alarms, they are often unable to quickly locate the root cause of the fault and perform targeted maintenance. It is even more difficult to automatically coordinate other gates or upstream water inflows after the anomaly occurs, resulting in the continuous amplification of the adjustment deviation caused by the fault, and then causing local or large-scale water supply interruptions or waterlogging disasters. If the operating data of the gate actuator cannot be collected in real time, accurately reduced in noise and finely analyzed, and the fault cannot be identified in time in the early stages of formation and corresponding fault-tolerant measures cannot be taken, the scheduling efficiency and water safety of the entire irrigation area will face major hidden dangers. Therefore, how to build a complete technical system covering multi-source data fusion, predictive fault diagnosis, automated fault tolerance and full-network collaborative scheduling to achieve accurate and reliable irrigation water supply under high equipment load and harsh working conditions has become a key problem to be solved by the present invention. Summary of the Invention

[0005] (1) Technical problems solved

[0006] In response to the shortcomings of the existing technology, the present invention provides an intelligent ditch dam adaptive adjustment method based on the Internet of Things. By focusing on the fault prevention and disposal of the dam actuator in farmland irrigation, it provides an overall method of multi-source data fusion, predictive diagnosis and automated fault tolerance, realizes closed-loop management of the dam actuator, and ensures the stable operation of large-scale farmland irrigation under load and outdoor environment, thereby solving the technical problems recorded in the background technology.

[0007] (2) Technical solution

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions: an intelligent ditch sluice dam adaptive adjustment method based on the Internet of Things, including: when an edge node receives a set trigger command, based on the raw data sequences collected by multiple sensors, wavelet denoising and time series alignment are used to generate multi-source fusion signal vectors and fault marker metadata in real time;

[0009] After receiving the fault tag metadata or predictive maintenance request from the previous stage, the system uses the multi-source fusion signal vector and historical maintenance records to assess the current risk and output a maintenance plan based on online health and fault severity, thus controlling the fault process before component damage occurs.

[0010] If the diagnosis results determine that the fault level is higher than expected and the secondary adjustment attempt is ineffective, the local controller will call the recovery capacity evaluation function and emergency linkage instructions based on the gate opening and upstream and downstream water level data, adjust the surrounding gates in real time, and generate a disposal record;

[0011] After the emergency linkage is completed and the complete fault event is recorded, the central dispatching center summarizes the multi-source fusion signal vector and the handling information of each gate, uses the objective function and adaptive entropy measurement to optimize the whole network dispatch and update the knowledge base.

[0012] Furthermore, multi-source sensors are arranged on each gate dam actuator, and wavelet transform is used to perform multi-scale denoising on each sensor signal to obtain the denoised signal. A fusion function is used to merge the denoising results of each sensor into a denoised multi-source fusion signal vector.

[0013] Furthermore, an abnormality scoring function is constructed based on the multi-source fusion signal vector to comprehensively evaluate the sensor signal and set a safety threshold. When the safety threshold is reached, the edge node generates a preliminary fault mark and records it in the local database, and solidifies the mark information into fault mark metadata on the edge side.

[0014] Furthermore, we acquire and integrate multi-source fusion signal vectors, fault marker metadata, and long-term accumulated operation logs, and introduce a comprehensive operating condition feature vector that characterizes the comprehensive operating condition information of the actuator:

[0015] After constructing the equipment health function, the typical characteristic patterns of various types of faults are obtained based on time series clustering and feature learning, and then the equipment health function is constructed; by analyzing and comparing the comprehensive operating condition feature vectors under different health states, the feature combination that is most sensitive to fault signs is extracted.

[0016] Furthermore, the online health level is calculated when the latest comprehensive operating condition feature vector is obtained. If the online health level shows a rapid decline or a persistent high fault risk, it is immediately compared with the fault tag metadata, and the fault severity is determined to form an online diagnosis result.

[0017] The comprehensive operating condition feature vector is compared with the historical typical fault patterns. When the match degree is higher than expected, the corresponding root cause location label and recommended maintenance measures are generated; and the fault severity is constructed. When the fault severity exceeds the set fault threshold, an urgent maintenance warning is simultaneously issued to the maintenance personnel or the dispatch center.

[0018] Furthermore, when the fault severity and online health or the fault tag metadata of the first step indicate that a certain actuator has a sudden fault, the local controller confirms and classifies the fault information, wherein:

[0019] The fault level is determined based on the unique fault tag and evaluation results, and the fault tolerance mechanism is activated accordingly:

[0020] If the fault level is at the mild or moderate stage, try to perform secondary adjustment actions and define a recovery capability evaluation function to quantify the self-recovery tendency of the actuator;

[0021] If the recovery capability evaluation function is lower than expected, the controller determines that it is currently in emergency mode and automatically enters the predefined linkage strategy; if the secondary adjustment action process successfully restores the fault indicator to a safe range, the fault is automatically recorded as a minor fault-successful self-correction event, and the result is transmitted back or reported to the cloud.

[0022] Furthermore, if the fault cannot be resolved in a short time or the fault level reaches a serious level, the local controller automatically initiates an emergency linkage. To achieve multi-gate coordination, optimization instructions are issued locally or by the upper control center, and a dynamic allocation function is introduced to represent the dynamic allocation of the openings of the relevant gates:

[0023] In emergency mode, a heuristic solution algorithm is used for the dynamic allocation function. After the solution is completed, a new gate scheduling instruction is immediately issued to the relevant execution agency; if the emergency linkage takes effect, the local controller will simultaneously generate the corresponding emergency event record and upload it to the cloud or central control terminal.

[0024] Furthermore, after completing fault tolerance and emergency linkage, each local controller encapsulates the fault event and its handling record into an event tag and transmits it back to the central control end;

[0025] The central control end constructs a global coordinated scheduling plan for the medium term or the next period based on the operating status of each gate, recent water demand, and the upstream water inflow and water level in the future period output by the water level and flow prediction model obtained by hybrid sequence prediction network training; and uses global optimization of the entire canal system on the central side. The optimal or near-optimal scheduling plan obtained will be sent to each gate and dam controller.

[0026] Furthermore, the entire process of fault and emergency response is recorded in a self-learning knowledge base. The knowledge base mainly includes: typical fault event IDs and cause analysis, fault detection and maintenance recommendation strategies, fault tolerance actions and emergency linkage processes, their effect evaluation, and subsequent impact on the entire channel system;

[0027] With the help of rule mining algorithms, the above information is summarized, and attempts are made to discover the correlation between fault types, treatment measures, and environmental conditions to form executable improvement rules.

[0028] Furthermore, under the self-learning framework, the central control end regularly iterates and updates the predictive maintenance model, fault diagnosis model and global scheduling algorithm parameters mentioned above, and introduces the generalization ability of the adaptive entropy measurement evaluation model, wherein: the entropy value difference between the adaptive entropy measurement and the previous iteration is used to evaluate the generalization ability of the model. Evaluate the model's discriminability changes in multiple scenarios:

[0029] like , keep the new parameters and continue fine-tuning;

[0030] like , then compare the sensitivity of each parameter to the entropy contribution and increase or decrease the corresponding value in the weight matrix or model parameter;

[0031] After completing several iterations and observing a continuously increasing entropy measure, the final parameter set and the corresponding weight matrix are verified and tested, and stored in the knowledge base together with the corresponding parameter adjustment strategy.

[0032] (3) Beneficial effects

[0033] The present invention provides an intelligent ditch dam adaptive adjustment method based on the Internet of Things, which has the following beneficial effects:

[0034] Based on the fusion collection and edge computing preprocessing of multi-source status data, the system can obtain and synchronize multi-source fusion signal vectors (such as motor current, gate opening, water level, etc.) in real time at the gate site and perform preprocessing such as wavelet denoising, greatly reducing network bandwidth occupancy and sensor noise interference, and improving the accuracy of early fault identification.

[0035] Based on in-depth analysis of predictive maintenance and fault diagnosis models, the system uses online health and fault severity to determine potential risks of actuators and generate maintenance recommendations, enabling equipment to be repaired before it is completely damaged. This significantly extends service life and reduces operation and maintenance costs. At the same time, it performs pattern recognition on comprehensive operating condition feature vectors (including external environmental characteristics) to avoid misjudgments based on single channels or manual experience.

[0036] By calculating a recovery capacity assessment function, the effectiveness of multiple opening and closing operations in clearing the blockage can be quickly determined. If the fault cannot be alleviated, the system automatically coordinates with surrounding gates to share downstream water supply or switch flood discharge channels, significantly shortening response time and improving resilience. The secondary regulation-emergency linkage mode effectively prevents the risk of widespread water outages or flooding caused by equipment failures. During this process, the local controller and the higher-level dispatch center share fault identification metadata and fault-tolerant decision-making results through high-speed communication, ensuring synchronized decision execution.

[0037] By integrating fault tolerance and emergency response experience, fault statistics, and multi-source fusion data into a unified knowledge base, we apply global scheduling objectives and adaptive entropy measures to optimize large-scale hydraulic scheduling and iterate models. This allows us to continuously enhance our adaptability to different gates and environmental scenarios, leveraging historical fault cases and real-time monitoring data, while continuously improving our diagnostic and scheduling strategies over time.

[0038] As a result, not only intelligent fault tolerance and real-time regulation are achieved at the local level, but also the irrigation scheduling plan is made more refined, reliable and scalable through self-learning iteration from the perspective of the entire network. Its significant beneficial effects are reflected in ensuring the continuity of agricultural water use, reducing the burden of manual inspections, improving equipment service life and preventing flood safety risks. At the same time, it can also comprehensively take into account energy costs, water use efficiency and maintenance investment, and realize the full-link upgrade of the gate and dam actuators from passive monitoring to active protection. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 The figure is a flow chart of the method for adaptively adjusting intelligent ditches, sluice gates and dams based on the Internet of Things according to the present invention. DETAILED DESCRIPTION

[0040] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0041] See also Figure 1The present invention provides an intelligent ditch dam adaptive adjustment method based on the Internet of Things, including:

[0042] Step 1: When the edge node detects a preset collection cycle or a cloud-based startup command, it uses feature processing methods such as wavelet denoising and time series calibration based on the raw data obtained by the sensor set to eliminate high-frequency interference and unify the sampling baseline. This generates high-quality data that integrates environmental factors such as water level and sediment, as well as preliminary fault markers, to promptly detect potential anomalies and store them in the edge database.

[0043] The step 1 includes the following:

[0044] Step 101: Sensor data acquisition and wavelet denoising fusion

[0045] Multi-source sensor deployment: Several types of sensors are deployed on each gate and dam actuator to form a sensor cluster. ,in:

[0046] Indicates the gate opening sensor, which is used to measure the actual opening and closing position of the gate; Represents the motor current sensor, which is used to monitor the load condition of the execution motor; Represents a vibration / temperature sensor used to assess whether a mechanical component has abnormal shock or overheating; Represents a water level sensor, used to monitor the upstream and downstream water level difference in real time; Indicates industrial camera image source (collecting sediment and weed distribution information), etc.

[0047] Since different sensors have different sampling frequencies and communication delays, the local edge computing nodes need to first unify the time axis. Represents the unified time mark inside the edge node, for each sensor In physical time The collected raw data can be expressed as ; In the edge node, it is mapped to , so that all sensor data can be synchronized at the same time top alignment;

[0048] In order to eliminate conventional mechanical vibration, electromagnetic interference and random pulse noise, wavelet transform is used to transform each sensor signal. Perform multi-scale denoising, is a pre-selected wavelet basis function (such as Meyer or Morlet), then the signal The discrete wavelet transform of can be expressed as:

[0049]

[0050] in: is an integer translation factor, which represents the translation amount in the time domain; is a positive real number scale parameter representing the resolution in the frequency domain;

[0051] is the total length of discrete sampling, that is, the number of sampling points in the current period; then the wavelet coefficients Perform threshold filtering to remove high-frequency or abnormal pulse items, and finally perform inverse wavelet transform to obtain the denoised signal ;

[0052] In order to use only purified reliable information in subsequent fault identification, predictive maintenance and other analyses, the fusion function is introduced Combine the denoising results of each sensor into the denoised multi-source fusion signal vector :

[0053]

[0054] For example, signals of different dimensions or forms can be integrated into a multi-source fusion signal vector according to the sensor type. :

[0055] , whose dimensions are , indicating that at time The latest observation values ​​of various sensors are stored in a unified data structure at the edge node.

[0056] When used, the multi-scale wavelet analysis method can more effectively retain fault precursors (such as slight current jitter and abnormal vibration spikes) than traditional simple filtering, while eliminating random noise; unified time marking of multi-source signals ensures that the subsequent modeling process can accurately associate various types of sensor information at the same time, avoiding data confusion in the time series dimension.

[0057] Step 102: Threshold detection and preliminary fault marking

[0058] After the wavelet denoising and fusion are completed, in order to timely detect possible fault signs (such as abnormal current rise and severe deviation), based on the multi-source fusion signal vector Building anomaly scoring function , for the moment Comprehensive evaluation of the sensor signal: define the denoised signal vector The time derivative of : ,

[0059]

[0060] Where: is the sensor signal vector after denoising and time synchronization; , represents the rate of change of the signal over time; Indicates the length of the integration window, in the interval The energy level of the internal cumulative quantized signal and its rate of change; for dimensional positive semidefinite matrix, which assigns weighting coefficients to the signal amplitude components; for dimensional positive semidefinite matrix, weighting the derivative terms (rate of change) of each signal; is a positive or non-negative real number scaling factor;

[0061] Set global or adaptive thresholds right Comparison is performed. If a limit is exceeded, a suspected fault flag is quickly triggered. The edge node immediately generates a preliminary fault flag and correlation ID for reference in the next step. The details are as follows:

[0062] Setting safety thresholds ,when When a fault occurs, the edge node immediately generates a preliminary fault tag and records it in the local database. The tag information is solidified as fault tag metadata on the edge side. , where the preliminary fault mark contains the following information:

[0063] Timestamp , involving sensor types and data values , Abnormality Score ,Subsequent association ID: a unique number can be assigned to this tag;

[0064] Therefore, the key data objects output by this step include: multi-source fusion signal vector after wavelet denoising and time alignment , preliminary fault labels and their metadata ;

[0065] When in use, with the help of the scoring mechanism, compared with the traditional mean-threshold or simple peak judgment mode, it can more flexibly capture moderate abnormal coupling of multiple channels, and complete preliminary fault marking at the edge node without waiting for cloud analysis, which greatly shortens the response time and improves the overall reliability of the system; by completing the threshold judgment and lightweight marking mechanism on the edge side, it reduces the communication burden for large-scale irrigation district automation systems and improves the level of detection intelligence.

[0066] Step 2: When the central control terminal receives the fault mark or predictive maintenance trigger signal uploaded by the edge node, it integrates the multi-source fusion information into the historical maintenance record and and fault severity Key indicators such as the wear trend of the actuator can be evaluated and preventive maintenance plans can be proposed, so that maintenance decisions can be made in time before the components become stuck and the risk of downtime can be reduced;

[0067] The second step includes the following:

[0068] Step 201: Historical pattern recognition and health model construction

[0069] Obtain and integrate the multi-source fusion signal vector from step 1 , fault tag metadata , and the long-term accumulated operation logs (including the actual gate opening, equipment maintenance records, etc.), which are stored in the central control terminal or the cloud database module, and the data representing the actuator at the time is introduced. The comprehensive working condition feature vector of the comprehensive working condition information :

[0070]

[0071] in is the multi-source fusion signal vector, Additional characteristics representing current environmental or external conditions (e.g. seasonal load levels, air temperature, silt load statistics, etc.);

[0072] In order to mine the degradation law and fault precursors of the actuator from historical data, the equipment health function is trained using historical data with normal / fault labels. , which can be considered as a mapping function of the actuator failure risk, where:

[0073] Based on time series clustering and feature learning (such as deep network or wavelet packet energy analysis), typical characteristic patterns of various faults are obtained, and then linear or nonlinear mapping relationships are used to evaluate health and define the equipment health function. :

[0074]

[0075] in: A set of feature segments extracted from historical data (which may include the time intervals before multiple typical failures) used to train or calibrate the health model; is a positive semidefinite matrix that determines the relative weights of different feature dimensions in the health status measurement. is a positive real number adjustment factor;

[0076] By analyzing the comprehensive working condition feature vectors under different health states (normal, mild wear, severe abnormality) Analyze and compare to extract the most sensitive feature combination for fault symptoms, and then complete fault diagnosis and risk assessment based on this model.

[0077] When used, by integrating the high-quality time series data obtained in the first step with the long-term historical maintenance records, data silos can be avoided and unified into the training process of the health model; by integrating the working condition feature vector The integration of environmental factors and multi-source signals can not only capture mechanical problems of the actuator, but also pay attention to external factors such as siltation and climate.

[0078] Step 202: Online fault location and intelligent early warning

[0079] In the device health function After completing offline training or calibration, it is deployed in the central control terminal or cloud platform for real-time operation. The comprehensive working condition characteristic vector Quickly calculate online health :

[0080]

[0081] If online health If the health level is rapidly declining or persistently in a high failure risk range, immediately mark the metadata with the failure Perform correspondence and comparison, determine the fault severity (such as minor abnormality, serious fault, etc.), and form online diagnosis results;

[0082] In order to further determine the specific fault source (such as motor bearing wear, gate blocking, communication abnormality, etc.), the comprehensive working condition feature vector is combined with pattern matching and expert system. Quickly compare with historical typical fault patterns:

[0083] When the matching degree is higher than expected, corresponding root cause location tags and recommended maintenance measures (such as reminders to clean the gate trough sediment, check the motor cooling system, etc.) are generated and shared with the maintenance department;

[0084] At the same time, the severity of the fault can be measured in the following ways :

[0085]

[0086] in: is the current health assessment value; The health baseline for normal working conditions or minor defects: is a positive real number magnification factor;

[0087] When the fault severity When the set fault threshold is exceeded, an urgent maintenance warning is sent to the maintenance personnel or dispatch center;

[0088] When in use, by connecting with the real-time data of the first step, continuous monitoring and dynamic updating of the health of the actuator can be achieved, which significantly improves the timeliness of fault detection. Based on the results of historical pattern matching and health models, potential lesions can be identified before the equipment is actually damaged, and proactive maintenance can be achieved to avoid large-scale shutdowns or irreversible failures. It can not only detect faults, but also give objective quantification of their causes and risk levels, thereby guiding the maintenance department to perform maintenance more efficiently and in a targeted manner.

[0089] Step 3: When the diagnosis result determines that the fault level is too high or the secondary adjustment fails to return to normal, the local controller will adjust the gate opening. Perform multiple checks with upstream and downstream water level data, then quickly initiate emergency linkage instructions and combine them with recovery capacity evaluation functions Evaluate the effectiveness of fault tolerance, adjust the opening of surrounding gates in real time, and cache fault tolerance records for subsequent use;

[0090] The step three includes the following:

[0091] Step 301: Fault-tolerance strategy triggering and secondary adjustment

[0092] When the fault severity of the second step and online health or the first step of the fault tag metadata If a sudden fault occurs in an actuator (such as an abnormal increase in motor current or a significant deviation in gate opening), the local controller will confirm and classify the fault information in the shortest possible time.

[0093] Based on the unique fault tag and evaluation results (such as online health or fault severity ), judge the fault level (minor, moderate, severe), and decide which fault tolerance mechanism to activate accordingly;

[0094] If the fault level is at a mild or moderate stage, the local controller will attempt to perform a secondary adjustment action, that is, execute the opening and closing instructions of the gate multiple times to eliminate problems such as small-scale blocking and short-term overcurrent; define the recovery capacity evaluation function Used to quantify the self-recovery tendency of the actuator:

[0095]

[0096] in: Indicates the real-time gate opening;

[0097] It is to execute the energy consumption model (or motor load function) to evaluate the energy consumption changes during multiple opening and closing actions. The specific form can be defined as:

[0098]

[0099] in, is the curve of gate opening (%) over time; , is the opening change rate (% / s);

[0100] is the motor current (A); is the power loss coefficient related to the motor resistance and efficiency;

[0101] is the nonlinear coefficient used to characterize mechanical transmission and hydraulic losses; is the observation window length of the secondary adjustment action, The time period for fault tolerance attempts.

[0102] Relative to the opening The gradient reflects the load of the door opening or closing action; is the short observation window (positive real number);

[0103] By evaluating the recovery capability function By analyzing the positive and negative and size of the gate, we can know whether the load gradually decreases after repeated opening and closing (indicating that the blocking resistance is reduced); if the recovery capacity evaluation function If the value is lower than expected, the controller will determine that it is in emergency mode and automatically enter the predefined linkage strategy;

[0104] If the secondary adjustment process is successful, the fault indicators (such as abnormality score, opening deviation, motor current and recovery ability evaluation function) ) is restored to a safe range, the fault is automatically recorded as a minor fault-successful self-correction event, and the result is transmitted back to step 2 and reported to the cloud.

[0105] When in use, in the early stage of the fault, short-term secondary adjustment is immediately performed without waiting for manual intervention, which can effectively solve minor faults caused by debris blocking the gate slideway and mud accumulation, etc. Dynamically capturing gate operation load changes can objectively judge the effectiveness of the fault-tolerant process.

[0106] Step 302: Emergency linkage and hydraulic dispatch optimization

[0107] If the fault cannot be corrected in a short period of time or reaches a critical level, the local controller will automatically initiate an emergency linkage. This includes: if the actuator gate fails, resulting in insufficient downstream flow, it will be necessary to link other parallel channels or diversion gates to open to maintain the established irrigation needs; if the gate is stuck in the open position and there is a risk of flooding, it will prioritize instructing the upstream main canal or adjacent sluice to reduce water flow, and may temporarily switch to flood discharge drainage channels;

[0108] In order to achieve multi-gate coordination, optimization instructions are issued locally or by the superior control center, and a dynamic allocation function is introduced to represent the dynamic allocation function of the opening of each related gate:

[0109]

[0110] in: Indicates the opening vector of each adjustable gate at the current moment (excluding failed gates or faulty actuators);

[0111] It is a hydraulic allocation objective function that integrates multiple conditions such as downstream water demand, channel flow constraints, and upstream safety limits; Represents constraint equations or inequalities (including gate opening range constraints, downstream flow upper and lower limit constraints, upstream water level safety constraints, etc.); is the feasible interval boundary of each constraint (such as safe water level, minimum water supply, etc.);

[0112] In emergency mode, a faster heuristic solution algorithm (such as hybrid linear constraint solving or greedy iterative algorithm) can be used for the dynamic allocation function, which is executed at the local or regional dispatch center. After the solution is completed, new gate dispatch instructions are immediately issued to the relevant execution agencies to ensure the safety of downstream water supply or upstream water level.

[0113] If the emergency linkage is effective, the local controller will simultaneously generate the corresponding emergency event record and upload it to the cloud or central control terminal, including: fault gate ID, fault type, timestamp emergency dispatch plan, that is, opening vector The final solution result; feedback on scheduling execution;

[0114] During operation, the system minimizes the impact of a single fault point on the irrigation system by synchronously adjusting the openings of surrounding gates and upstream water inflow. This demonstrates the solution's rapid response capability in overall hydraulic scheduling. In the event of a fault, it can flexibly switch between safety-first and water-first strategies, effectively preventing the risk of flooding or drought caused by gate failure.

[0115] Through the dynamic scheduling equation min By considering multiple hydraulic constraints in parallel and elevating automated fault tolerance to network-level optimization, the system breaks through the limitation of traditional irrigation systems that can only respond passively to a single gate. When an actuator fails or faces high risk, automated response measures can be quickly taken to minimize the impact on the overall operation of the irrigation system. Not only can minor faults be quickly eliminated, but in severe fault scenarios, other gates can also be efficiently directed to coordinate diversion or protection, ultimately achieving the goal of moving the entire irrigation district management from passive emergency response to self-learning intelligent operation and maintenance.

[0116] Step 4: When the local emergency linkage and fault tolerance process is completed and a global event record is generated, the central dispatch center will combine the fault processing metadata with the multi-source fusion signal vector , fault severity and online health Merge the information into the whole canal fault data set , through the objective function and adaptive entropy measure Execute network-wide scheduling parameter iteration, synchronously update the knowledge base and issue new opening plans, improving the prediction accuracy and fault tolerance of the intelligent irrigation system;

[0117] The step 4 includes the following contents:

[0118] Step 401: Multi-source data transmission and global coordinated scheduling

[0119] After completing fault tolerance and emergency linkage, each local controller encapsulates the fault event and its handling record (including the final opening scheduling vector, fault recovery time, emergency process start trigger time, etc.) as an event tag and transmits it back to the central control end;

[0120] At the same time, the raw sensor data accumulated in the edge or cloud in the first and second steps, and the multi-source fusion signal vector after denoising , fault severity , and the fault-tolerant action records formed in the third step are collected and integrated by the central database to form a global level full channel fault data set ;

[0121] After completing the data aggregation, the central control terminal constructs a global coordinated scheduling plan for the medium term or the next period based on the operating status of each gate, the recent water demand (such as irrigation season, farmland area, and crop type), and the upstream water inflow and water level in the future period output by the water level and flow prediction model obtained by hybrid sequence prediction network training.

[0122] Based on the real-time operating status of the entire canal system, future water demand, and water level and flow forecasts, the central control terminal first integrates information such as the current opening of each gate, upstream and downstream water levels, irrigation season, field area, and crop water requirements as input variables, and uses hydrological models to predict water inflow and water level changes in future time periods. Then, within a limited time domain, using the gate opening trajectory as the decision variable, an optimization function is constructed that includes multiple objectives such as minimizing downstream water supply deviation, avoiding over-limit water release, and controlling upstream water level risks, combining the opening, flow, and water level constraints. The model predictive control (MPC) framework and numerical optimization algorithm are then used to determine the optimal gate opening sequence within the rolling time period. Only the initial command is issued and it is continuously recalculated based on the latest status and forecast results. In this way, a global coordinated scheduling plan that takes into account both water supply accuracy and safety can be dynamically generated and implemented.

[0123] To improve the accuracy of scheduling decisions, a global optimization of the entire channel system is performed on the central side as follows:

[0124]

[0125] in: Represents the target opening matrix (or a set of control intervals) of all controllable gates in the entire irrigation area within a certain time sequence in the future;

[0126] To achieve the overall water conservancy optimization goal, multiple factors such as the stability of downstream water demand, upstream water level safety redundancy, and regional electricity costs can be comprehensively considered; A set of vectorized inequalities or equations to satisfy hydrological constraints, safety constraints, and equipment status constraints; is the threshold interval allowed by the corresponding constraint; Indicates that the elements are less than or equal to each other item by item (suitable for multi-dimensional constraint scenarios);

[0127] The optimal or near-optimal scheduling plan obtained will be sent to each dam controller to guide the gate opening strategy in subsequent normal cycles (such as daily / weekly); after integrating the fault-tolerant compensation method when a failure occurs in the third step, even if individual gates fail again, irrigation and safety can be maximized at the network level.

[0128] When in use, the fault, sensor and scheduling data generated in the previous steps are systematically classified through a complete indexing and labeling mechanism, laying a data foundation for subsequent self-learning. Different from the third step which focuses on single-point emergency response, this step makes overall operations at the central system level, which greatly improves the water use efficiency and safety redundancy of the irrigation area. At the same time, compared with the traditional irrigation system which only relies on manual experience or segmented scheduling, this step uses a multi-dimensional objective optimization function to optimize the system. Integrating various elements and linking them with fault-tolerant mechanisms can achieve a balance between efficiency and safety.

[0129] Step 402: Self-learning knowledge base construction and continuous improvement

[0130] After completing the data merging and global scheduling in step 401, the entire process of fault and emergency handling is recorded and included in the self-learning knowledge base. The knowledge base mainly includes:

[0131] Typical fault event IDs and cause analysis, fault detection and maintenance recommendations, fault-tolerant actions and emergency linkage processes (determining success or failure), their effectiveness evaluation, and subsequent impact on the entire canal system (whether further diversion or scheduling is necessary). Rule mining algorithms are used to summarize this information, attempting to identify correlations between fault types, treatment measures, and environmental conditions, and developing executable improvement rules (e.g., increasing the frequency of gate channel cleaning during high sediment seasons, requiring early maintenance of certain motor models when the temperature exceeds a threshold).

[0132] Under the self-learning framework, the central control end regularly (or on demand) iteratively updates the aforementioned predictive maintenance model, fault diagnosis model, and global scheduling algorithm parameters, gradually adapting them to the latest actual operating data and fault handling experience.

[0133] In order to further accelerate the self-learning process and take into account the advancedness, the adaptive entropy measure is introduced Evaluate the generalization ability of the model, where: Represents model parameters In a given scenario (which can be understood as a fault scenario or environmental state), The predicted probability distribution vector of the class fault type (or classification label);

[0134] In the scene domain Integrate the predicted probability distribution vector and introduce a weight matrix between the vector and the logarithmic vector , define the adaptive entropy measure as follows:

[0135]

[0136] in: The parameter vector (or parameter matrix) representing the model, such as the weight of the device health function, the hyperparameters of the fault diagnosis network, the priority coefficient of the global scheduling algorithm, etc.

[0137] Indicates the model parameters and scenes Next, right The predicted probability of a fault (or classification label) is , and each component is in , and the sum of all components is 1; An index of scenarios or environmental conditions, which may include different failure modes, seasonal hydrological conditions, and different equipment status combinations; for The positive semidefinite matrix is ​​used to assign different weights to each fault category in the measurement;

[0138] According to the adaptive entropy measure , you can iteratively adjust parameters (such as increasing or decreasing the weight of certain features) to continuously optimize the model and more accurately distinguish fault scenarios. You can also consolidate effective optimization solutions into a knowledge base to provide automated configuration for subsequent deployments, as follows:

[0139] According to the entropy value and the entropy value of the previous iteration Evaluate the model's discriminability changes in multiple scenarios:

[0140] like , which shows that the new parameters have better generalization ability in fault differentiation and can be retained and further fine-tuned;

[0141] like , then compare the sensitivity of each parameter to the entropy contribution (which can be obtained through numerical gradient or parameter sensitivity analysis), and in the weight matrix or model parameters Increase or decrease the corresponding value in order to improve the responsiveness to low entropy scenarios;

[0142] After completing several iterations and observing a continuous increase in the entropy measure, the final parameter set With the corresponding weight matrix Conduct validation testing to ensure they demonstrate improved class discrimination capabilities across both historical and emerging fault scenarios.

[0143] The winning parameter set, along with the corresponding parameter tuning strategy, is then stored in a knowledge base, and an automated configuration script is generated for rapid loading and application in newly deployed edge controllers or central models without manual intervention.

[0144] By combining information entropy to evaluate the model's ability to distinguish between different fault scenarios, larger values ​​generally indicate that the model has higher sensitivity and adaptability to various fault types. Based on this entropy measure, iterative parameter adjustments (such as increasing or decreasing the weights of certain features) are adopted to continuously optimize the model and achieve more accurate differentiation of fault scenarios. Optimization solutions with significant results can be consolidated into a knowledge base to provide automated configuration for subsequent deployment.

[0145] During use, each fault and handling information can be accumulated and reviewed through the knowledge base to avoid repeating the same mistakes and provide a reference for future equipment upgrades or promotion in multiple irrigation areas. The introduction of information entropy measurement to quantify the diagnosis and optimization effects of the model can enable the entire system to self-iterate in actual applications and improve the adaptability of the scene. Traditional irrigation automation mostly stays in static experience or manual parameter tuning. Here, by introducing the knowledge base The information entropy method adaptively strengthens the model, combining expert experience with the advantages of big data mining, making it more targeted.

[0146] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0147] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0148] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is only for some logical functions. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0149] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0150] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. An intelligent adaptive adjustment method for ditch gates and dams based on the Internet of Things, characterized by: include, If the edge node receives a set trigger command, it uses wavelet denoising and time series alignment based on the raw data sequences collected by multiple sensors to generate multi-source fusion signal vectors and fault mark metadata in real time; After receiving the fault tag metadata or predictive maintenance request from the previous stage, the system uses the multi-source fusion signal vector and historical maintenance records to assess the current risk and output a maintenance plan based on online health and fault severity, thus controlling the fault process before component damage occurs. If the diagnosis result determines that the fault level is higher than expected and the secondary adjustment attempt is invalid, the local controller calls the recovery capacity evaluation function and emergency linkage instructions based on the gate opening and upstream and downstream water level data, adjusts the surrounding gates in real time and generates a disposal record; among them, the recovery capacity evaluation function is defined Used to quantify the self-recovery tendency of the actuator: ;in: Indicates the real-time gate opening; It is an execution energy consumption model to evaluate the energy consumption changes during multiple opening and closing actions. The specific form is defined as: ;in, is the curve of gate opening over time; , is the opening change rate; is the motor current; is the power loss coefficient related to the motor resistance and efficiency; is the nonlinear coefficient used to characterize mechanical transmission and hydraulic losses; is the observation window length of the secondary adjustment action, The time period for error tolerance attempts; Relative to the opening The gradient reflects the load of the door opening or closing action; is the short observation window (positive real number); After the emergency linkage is completed and the complete fault event is recorded, the multi-source fusion signal vector and the gate handling information are summarized, and the objective function and adaptive entropy measurement are used to optimize the whole network scheduling and update the knowledge base. Integrate the predicted probability distribution vector and introduce a weight matrix between the vector and the logarithmic vector , define the adaptive entropy measure as follows: ;in: The parameter vector representing the model, including the weight of the device health function, the hyperparameters of the fault diagnosis network, and the priority coefficient of the global scheduling algorithm; Indicates that under the model parameters and scenarios, The predicted probability of a fault is , and each component is in , and the sum of all components is 1; An index of scenarios or environmental conditions, including different failure modes, seasonal hydrological conditions, and different equipment status combinations; for The positive semidefinite matrix is ​​used to assign different weights to each fault category in the measurement.

2. The method for adaptively adjusting intelligent ditch dams based on the Internet of Things according to claim 1 is characterized by: Multi-source sensors are arranged on each gate and dam actuator, and wavelet transform is used to perform multi-scale denoising on each sensor signal to obtain the denoised signal; The fusion function is used to merge the denoising results of each sensor into a denoised multi-source fusion signal vector.

3. The method for adaptively adjusting intelligent ditch dams based on the Internet of Things according to claim 2 is characterized by: An anomaly scoring function is constructed based on the multi-source fusion signal vector, the sensor signal is comprehensively evaluated, and a safety threshold is set; when the anomaly score exceeds the safety threshold, the edge node generates a preliminary fault mark and records it in the local database, and the mark information is solidified as fault mark metadata on the edge side.

4. The method for adaptively adjusting intelligent ditch dams based on the Internet of Things according to claim 3 is characterized by: Acquire and integrate multi-source fusion signal vectors, fault marker metadata, and long-term accumulated operation logs, and introduce comprehensive operating condition feature vectors that characterize the comprehensive operating condition information of the actuator; After constructing the equipment health function, the typical characteristic patterns of various faults are obtained based on time series clustering and feature learning, and then the equipment health function is constructed.

5. The method for adaptively adjusting intelligent ditch dams based on the Internet of Things according to claim 4 is characterized by: When the most recent comprehensive operating condition feature vector is obtained, the online health level is calculated. If the online health level shows a rapid decline or a persistent high fault risk, it is compared with the fault marker metadata, and the fault severity is determined to generate an online diagnosis result. The comprehensive operating condition feature vector is compared with the historical typical fault patterns. When the match degree is higher than expected, the corresponding root cause location label and recommended maintenance measures are generated; and the fault severity is constructed. When the fault severity exceeds the set fault threshold, an urgent maintenance warning is simultaneously issued to the maintenance personnel or the dispatch center.

6. The method for adaptively adjusting intelligent ditch dams based on the Internet of Things according to claim 5 is characterized by: When the fault severity and online health or the fault tag metadata in the first step indicate that an actuator has a sudden fault, the local controller confirms and classifies the fault information, including: The fault level is determined based on the unique fault tag and evaluation results, and the fault tolerance mechanism is activated accordingly: If the fault level is mild or moderate, a secondary adjustment action is attempted and a recovery capability evaluation function is defined to quantify the self-recovery trend of the actuator. If the recovery capability evaluation function is lower than expected, the controller determines that it is currently in emergency mode and automatically enters a predefined linkage strategy.

7. The method for adaptively adjusting intelligent ditch dams based on the Internet of Things according to claim 6 is characterized by: If the fault cannot be resolved within the scheduled time or the fault level reaches a serious level, the local controller automatically initiates emergency linkage, and optimization instructions are issued locally or by the superior control center. A dynamic allocation function is introduced to represent the dynamic allocation of the openings of each relevant gate. In emergency mode, a heuristic solution algorithm is used for the dynamic allocation function. After the solution is completed, a new gate scheduling instruction is immediately issued to the relevant execution agency.

8. The method for adaptively adjusting intelligent ditch dams based on the Internet of Things according to claim 7 is characterized by: After completing fault tolerance and emergency linkage, each local controller encapsulates the fault event and its handling record into an event tag and transmits it back to the central control end; The central control end builds a global coordinated scheduling plan for the medium term or the next period based on the operating status of each gate, the recent water demand, and the upstream water inflow and water level in the future period. The center uses global optimization of the entire canal system, and the optimal or approximately optimal scheduling plan obtained is sent to each gate and dam controller.

9. The method for adaptively adjusting intelligent ditch dams based on the Internet of Things according to claim 8, characterized in that: The entire process of fault and emergency response is recorded in a self-learning knowledge base. This knowledge base includes: typical fault event IDs and cause analysis, fault detection and maintenance recommendations, fault-tolerant actions and emergency linkage processes, their effectiveness evaluation, and subsequent impact on the entire channel system. With the help of rule mining algorithms, the above information is summarized, and attempts are made to discover the correlation between fault types, treatment measures, and environmental conditions to form executable improvement rules.

10. The method for adaptive regulation of intelligent ditch dams based on the Internet of Things according to claim 9, characterized in that: Under the self-learning framework, the central control end regularly iterates and updates the predictive maintenance model, fault diagnosis model and global scheduling algorithm parameters mentioned above, and calculates the entropy value difference between the adaptive entropy measure and the previous iteration. Evaluation model's discrimination changes in multiple scenarios: If , keep the new parameters and continue fine-tuning; if , then compare the sensitivity of each parameter to the entropy contribution and increase or decrease the corresponding value in the weight matrix or model parameter; After completing several iterations and observing a continuously increasing entropy measure, the final parameter set and the corresponding weight matrix are verified and tested, and stored in the knowledge base together with the corresponding parameter adjustment strategy.

Images